Utilizing network features to detect erroneous inputs

Neural networks are vulnerable to a wide range of erroneous inputs such as corrupted, out-of-distribution, misclassified, and adversarial examples. Previously, separate solutions have been proposed for each of these faulty data types; however, in this work I show that the collective set of erroneous inputs can be jointly identified with a single model. Specifically, I train a linear SVM classifier to detect these four types of erroneous data using the hidden and softmax feature vectors of pre-trained neural networks. Results indicate that these faulty data types generally exhibit linearly separable activation properties from correctly processed examples. I am able to identify erroneous inputs with an AUROC of 0.973 on CIFAR10, 0.957 on Tiny ImageNet, and 0.941 on ImageNet. I experimentally validate the findings across a diverse range of datasets, domains, and pre-trained models.

Name thesis:degree_name

Master of Science (M.S.)

Level thesis:degree_level

Masters

Discipline thesis:degree_discipline

Computer Science

Grantor dc:publisher

Colorado State University. Libraries

Year dc:date.issued

2020

Authors dc:creator

• Gorbett, Matthew, author
• Blanchard, Nathaniel, advisor
• Anderson, Charles W., committee member
• King, Emily, committee member

Statement dc:rights

• Copyright and other restrictions may apply. User is responsible for compliance with all applicable laws. For information about copyright law, please see https://libguides.colostate.edu/copyright.

Language dc:language.iso

eng, English

Identifier URI

https://doi.org/10.25675/3.02487

OAI identifier oai:identifier

oai:mountainscholar.org:10217/219573